Weekly GEO research: why one platform label is not enough

AI-visibility measurement often begins with a convenient but overly coarse row: platform, prompt, answer. That format hides conditions that change the source set. One operator may offer ordinary search, deep research, pro mode, voice, app surfaces, and different models. Google may expose different surfaces and market rollouts. The observation unit should therefore describe a specific run: surface, mode, model, device, language, region, and time, not just a platform brand. Without these fields, weekly comparisons mix product change with experiment change.

Section sources:[1] Perplexity[2] ACL Anthology[3] Google Analytics

Perplexity’s Pro Search description shows that focus, selected model, and dialogue mode can be part of the answer condition. This does not establish a ranking mechanism or show that one mode is better for every brand. It does require honest labelling: Pro Search should not be pooled with ordinary search without a qualification. The run log should preserve mode, selected model, source focus, duration, and whether the dialogue continued. A rerun in a fresh context is a different condition, not a technical duplicate.

Section sources:[1] Perplexity

Primary studies of generative search show that source sets can differ substantially across systems and repeated runs. This does not mean answers are random or equally unreliable. It means that one attractive screenshot is weak evidence. Each cluster needs repeated runs, fixed conditions, and preserved answers. A visibility change should not be announced when it is smaller than normal repeat noise. The sufficient-sample threshold should depend on citation-bearing responses and a stated stopping rule, not on a universal prompt count.

Section sources:[2] ACL Anthology

Google Analytics documentation matters for traffic interpretation: AI Overviews and AI Mode may remain under Organic Search instead of a separate AI Assistants channel. A single AI-referral report is therefore likely incomplete. Store source and medium, landing page, search referrer, Telegram campaigns, and direct visits separately. When a source provides no referrer, that is not proof of AI origin. The analytics layer should show uncertainty and must not turn channel classification into a fact about what influenced a user.

Section sources:[3] Google Analytics

Paid presence and an organic citation answer different questions even when they appear in one AI interface. A public report should give advertising its own surface, date, query set, and classification rule. It must not be folded into citation rate or a blended score: buying an impression does not prove source selection, while an organic link says nothing about the advertising campaign. If one domain appears both in an ad and among sources, those are two observations with different meanings and denominators.

A minimal schema for future studies includes run_id, operator, surface, mode, model, device, language, region, prompt, repeat, answer text, source set, mention, citation, recommendation, paid presence, referral, and date. Every classification needs a rule and correction history. This issue has no production logs from closed systems, so retrieval remains unavailable; research findings belong to their own corpora, and GA4 behavior belongs to a specific property configuration. The defensible conclusion is to improve the protocol, not publish a universal platform ranking.

Section sources:[1] Perplexity[2] ACL Anthology[3] Google Analytics

Each issue should preserve not only its final thesis but also the record of how that thesis was formed. Useful fields include check date, source version, exact observation, editorial classification, and the reason for a change from the previous week. If a source disappears, changes text, or changes rollout, the old state should not be silently overwritten. Preserve the public URL and describe what changed. This record lets readers distinguish new information from a new interpretation and reduces the risk of turning a note into automated freshness theatre.

Section sources:[1] Perplexity[2] ACL Anthology[3] Google Analytics

To compare weekly results, every record should begin with a run passport. It should state the exact question, interface language, country, device, surface, mode, selected model, and repeat number. Preserve the complete answer with its original links, then label events separately: source found, object named, link shown, and recommendation given. If the interface changed or part of the answer is unavailable, record that as a limitation rather than filling the gap with a guess. For an editor, this card is more useful than a global score: it shows which claims can be published, which need replication, and which must remain hypotheses. Several cards from one week form a reproducible corpus that can be updated without losing context or mixing different products under one platform name.

Section sources:[1] Perplexity[2] ACL Anthology[3] Google Analytics

For replication, read this issue as a measurement protocol rather than a platform ranking. Before comparing runs, fix the launch time, model, surface, device, and query set, then preserve the original answer and its links. Record separately what was observed and what the editorial team infers. If the mode or region changes, create a new observation row instead of merging it with the earlier run. This lets an editor explain why two similar checks diverged and which differences belong to the environment rather than to page quality.